Skip to main content
This page covers two things: making a voice say a text (the “Voice generator” tool), and creating a custom voice (the “My voices” page).

​
Making a voice say a text

You give the text to say, or you ask the assistant to write it. It launches the synthesis with two pieces of information: the exact text, with no stage directions, and a one sentence tone direction. It picks the voice from the list, or you choose it yourself in “Voice” (“My voices”, catalog).
A text that is too long is cut by the assistant into several generations. To have the text written: “Write and voice a two sentence hook for a cooking video”.

​
Creating a voice from a description

In “My voices”, “Describe a voice” takes a name (60 characters) and a description (500 characters). Say the gender, the approximate age, the timbre, the grain, the pace, the accent.
Example provided by the app: “Deep, calm voice of a 50 year old man, slightly husky, southern accent”. A variant offers three voices to listen to before keeping the one you like. It accepts a longer description (1,000 characters): add the role (calm storyteller of true stories) and the acting (pauses before each reveal). Each batch of three proposals costs a few credits, and so does the voice you keep.
You can also clone your own voice by recording a sample: it is your voice, or the voice of a person who gave you their consent. Custom voices depend on your plan and your quota.

​
Common mistakes

  • Putting a stage direction in the text (“(he laughs)”): it would be read aloud. Put the acting in the tone direction.
  • A tone in several paragraphs: one sentence is enough.
  • Describing the voice with the emotion of a single line: describe the voice, not what is said.
  • Naming a real or famous person: describe the timbre.
  • Writing in one language and choosing a voice from another: set “Language” and pick a voice that speaks it.

​
Settings that go with it